Predictors of Survival in Patients with Sarcoma Admitted to the Intensive

Gupta et al. Clin Sarcoma Res (2016) 6:12 DOI 10.1186/s13569-016-0051-5 Clinical Sarcoma Research

RESEARCH Open Access Predictors of survival in patients with sarcoma admitted to the intensive care unit Rohan Gupta1, Neda Heshami1, Chouhan Jay1, Naveen Ramesh2, Juhee Song3, Xiudong Lei3, Erfe Jean Rose4, Kristen Carter5, Dejka M. Araujo5, Robert S. Benjamin5, Shreyaskumar Patel5, Joseph L. Nates4 and Vinod Ravi5*

Abstract Background: Advances in treatment of sarcoma patients has prolonged survival but has led to increased disease- or treatment-related complications resulting in greater number of admissions to the intensive care unit (ICU). Survival and long-term outcome information about such critically ill patients with sarcoma is unknown. Methods: The primary objective of the study was to determine the ICU and post-ICU survival rates of critically ill sarcoma patients. Secondary objectives included determining the modifiable and non-modifiable predictors of poor survival. We performed a retrospective chart review of sarcoma patients admitted to the ICU at The University of Texas MD Anderson Cancer Center between January 1, 2005, and December 31, 2012. Main outcome measures were ICU mortality, in-hospital mortality and 1, 2, and 6-month survival rates. Covariates such as histological diagnosis, disease characteristics, chemotherapy use, Charlson comorbidity index, Sequential Organ Failure Assessment (SOFA) scores, and clinical findings leading to ICU admission were analyzed for their effects on survival. Results: We identified 172 admissions over the 8-year study period hat met our inclusion criteria. The study population was 45.9 % males with a median age of 52 years. The most common sarcoma subgroups were high-grade unclassified sarcoma (25 %) and bone tumors (17.4 %). The ICU mortality rate was 23.3 % (95 % confidence interval [CI], 16.9–29.6 %), and an additional 6.4 % of patients died before hospital discharge (95 % CI, 22.9–37.1 %). 6-month OS rates were 41 %. The median SOFA scores on admission were 6 (inter quartile range (IQR), 3.5–9) in ICU survivors and 10 (IQR, 6.5–14) in ICU non-survivors. Increase in SOFA scores 6 led to poor outcomes (ICU survival 13.3 %, OS 6.7 %). Charlson comorbidity index (HR 1.139, 95 % CI 1.023–1.268, p≥ 0.02) and discharge SOFA scores (HR 1.210, 95 % CI 1.141–1.283, p < 0.0001) correlated with overall survival. = Conclusions: Our results suggest that patients that are admitted to the ICU for respiratory failure, cardiac arrest, septic shock, acute renal failure or acidosis and also have a high SOFA score with subsequent worsening in the ICU have poor prognosis. Based on the retrospective data which needs further validation we can recommend that judicious approach should be taken in patients with predictors of poor survival before subjecting them to aggressive treatment. Keywords: Cancer, Sarcoma, ICU, Survival, SOFA

*Correspondence: [email protected] 5 Department of Sarcoma Medical Oncology, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd # 450, Houston, TX 77030, USA Full list of author information is available at the end of the article

© 2016 The Author(s). This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 2 of 12

Background of patients admitted to the ICU with sarcoma during the Sarcomas are a rare, histologically and behaviorally study period. Secondary outcomes were in-hospital mor- diverse group of malignant connective tissue tumors that tality and 1, 2, and 6-month survival rates, which were make up approximately 1 % of all adult malignancies and defined by similar means. Overall survival (OS) for each 12 % of pediatric cancers [1, 2]. As a result of the substan- patient was measured from the time of initial ICU admis- tial progress made in the past 2 decades in understand- sion to the last date of contact or death. We defined ICU ing the behavior and molecular pathogenesis of sarcoma, survival as short-term or acute survival and 6-month sur- new therapies have been developed. Advances in treat- vival as long-term survival. Median follow-up time was ment may prolong survival but can lead to increased calculated from the date of ICU admission. disease- or treatment-related complications requiring Patient’s current and previous chemotherapy regimens aggressive critical care. Due to the rarity of this class of were examined to evaluate the impact of various chemo- tumors, the survival of critically ill sarcoma patients has therapy treatments on survival. Tumor burden was noted not been well studied. by recording the various sites of metastasis, such as head The treatment of such patients requires a multidisci- and neck, musculoskeletal, heart, lung, liver, gastrointes- plinary approach, with coordination among oncologists, tinal tract, and spleen. The Charlson comorbidities index critical care physicians, consulting services, ancillary (CCI) was used to assess the role of serious comorbid dis- staff, and patients’ families. Despite the availability of ease in the survival of patients in the study [9]. advanced life support devices in intensive care centers in Clinical findings present at the time of ICU admis- the United States, it is difficult for both patients and phy- sion were recorded to assess acute illness (definitions in sicians to objectively determine the effect of such heroic parentheses). Variables recorded included acute renal measures on prognosis or quality of life. Studies up to the failure, anemia, thrombocytopenia, pancytopenia, 1990s had shown that among all diseases, patients with hemorrhage (including bleeding from gastrointestinal cancer had the lowest intensive care unit (ICU) survival tract), lactic acidosis, heart failure, pulmonary embo- rates, and the majority of these patients died soon after lism, respiratory failure (use of noninvasive or invasive their hospital discharge [3–5]. However, over the past mechanical ventilation), cardiac arrest, atrial fibrillation, two decades, the approach to treatment in such cases cardiac dysrhythmia, pneumonia, septic shock, hypoten- has been shifting. Some studies have shown that patients sion (systolic blood pressure <100 or on vasopressors), lacking predictors of poor survival outcomes are consid- hypertension, neutropenic fever (absolute neutrophil ered as good candidates for aggressive therapy [6–8]. count <1500), altered mental status, and malnutrition The primary objective of our study was to determine (patient on feeding tube or total parental nutrition). the ICU and post-ICU survival rates of critically ill sar- Severity of illness at the time of ICU admission was coma patients. Secondary objectives included determin- measured by Sequential Organ Failure Assessment ing the modifiable and nonmodifiable predictors of poor (SOFA) scores. Maximum SOFA scores and discharge survival. SOFA scores were also obtained to track the patients’ progress while they were in the ICU. Change in SOFA Methods scores were calculated by subtracting the admission Patient population SOFA scores from the maximum SOFA scores. Organ After obtaining institutional review board approval, we failure at the time of ICU admission was determined by retrospectively reviewed the electronic medical records an admission SOFA score ≥2 per organ system. The total of 212 critically ill patients with sarcoma who had been number of organ failures was calculated for each patient. admitted to the ICU at The University of Texas MD Anderson Cancer Center between January 1, 2006, and Analysis December 31, 2012. For the purpose of this analysis, we Patient characteristics were tabulated and compared limited the study population to patient’s first ICU admis- between groups by using the Chi square test or Fisher sion only. We also excluded from the study population exact test as appropriate for categorical variables and by patients who had been admitted to the ICU for periop- the nonparametric Wilcoxon rank sum test for continu- erative care. ous variables. A multivariate logistic regression model was fitted to examine the relationship between death in Study design the ICU and clinical characteristics. Patients who were This retrospective study was designed to identify predic- lost to follow-up or alive were censored at their dates of tors of poor survival that contribute to ICU mortality. last contact. The Kaplan–Meier product limit method was ICU mortality was defined as the percentage of patients used to estimate the survival outcomes of all patients by with sarcoma who died in the ICU among total number groups; the log-rank statistic was used to compare groups. Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 3 of 12

Cox proportional hazards models were fitted to deter- non-survivors. Patients with admission SOFA scores of mine the association of patient and clinical characteristics 11 or more had a lower ICU survival rate than did those with OS. Variables that had significant univariate log- with scores less than 11 (45.7 vs 80 % or more) (Table 2). rank p values were candidates for the multivariate model. An increase of 6 or more in the SOFA score from the time Results were expressed in hazard ratios (HRs), odds ratios of admission significantly affected short- (ICU survival (ORs) and 95 % confidence intervals (CIs). p values of less 13.3 %) and long-term outcomes (OS 6.7 %) (Table 3). than 0.05 were considered statistically significant; all tests Multivariate logistic regression model for death in the were two-sided. Statistical analyses were carried out by ICU showed that SOFA admission score (OR 1.23, 95 % using SAS 9.4 (SAS Institute Inc., Cary, NC) and S-Plus CI 1.12–1.35, p < 0.0001) was associated with death 8.2 (TIBCO Software Inc, Palo Alto, CA, USA). in the ICU. Variables that were initially included in the model and then reduced in a stepwise selection were Results SOFA admission score and number of metastatic sites. Patient characteristics We identified a total of 212 sarcoma admissions to the OS from ICU admission ICU at MD Anderson between January 1, 2005, and Median follow-up among all patients was 2.9 months December 31, 2012. We excluded 23 ICU admissions of (range 0.02–77.5 months). At the time of this analysis, patients who were admitted to the ICU multiple times 135 patients (78.5 %) had died. For the whole cohort, 1, 2, during the study course. Of the remaining 189 admis- and 6-month OS rates were 64, 57, and 41 %, respectively sions, 17 perioperative admissions were excluded, leav- (Table 4). The Kaplan–Meier curves for OS are shown in ing a sample of 172 first-time ICU admissions. The study Fig. 1. population was 45.9 % male with median age of 52 years In multivariable Cox proportional hazards models (interquartile range [IQR] 38–62 years) (Supplemental (Table 5), patients with gastrointestinal stromal tumors Digital Content—Table 1). The most common sarcoma (HR 0.281, 95 % CI 0.119–0.662, p = 0.004) and leio- subgroups were unclassified high-grade sarcoma (25 %), myosarcoma (HR 0.375, 95 % CI 0.160–0.880, p = 0.02) bone sarcoma (Ewing sarcoma, osteosarcoma, and chon- had a lower risk of death than did patients with unclas- drosarcoma; 17.4 %), vascular sarcoma (angiosarcoma sified high-grade sarcoma. Patients with higher Charl- and epithelioid hemangioendothelioma; 9.9 %), and leio- son Comorbidity Index (HR 1.139, 95 % CI 1.023–1.268, myosarcoma (7.6 %). The ICU mortality rate was 23.3 % p = 0.02) and those with higher SOFA scores at discharge (95 % CI 16.9–29.6 %), and the hospital mortality rate (HR 1.210, 95 % CI 1.141–1.283, p < 0.0001) also had was 29.7 % (95 % CI 22.9–37.1 %). The median Charlson higher risk of death (Table 5). Median survival rates were comorbidity index was 6 (IQR 6–7) owing to presence of lower in patients with acute renal failure, cardiac arrest, metastatic cancer in most of the patients at the time of pneumonia, septic shock, and respiratory failure. admission. Discussion Death in the ICU Decisions about the intensive care treatment of critically There were 40 patient deaths (23.3 %) in the ICU. In the ill cancer patients with poor prognoses are challenging univariate logistic regression model for the death at the and need to be evaluated on a patient-by-patient basis. ICU, short-term mortality did not correlate with tumor Our study benefits both physicians and family members site, histology, disease status, presence of metastatic dis- by providing objective data on ICU mortality, long-term ease, Charlson comorbidity index or neutropenia at the survival, and objective predictors of survival. Our results time of ICU admission. However, patients with clinical showed that among sarcoma patients, the ICU mortal- findings of acidosis, acute renal failure, cardiac arrest, ity, in-hospital mortality, and long-term survival rates hypotension (including septic shock), pneumonia, septic were 23.3, 29.7, and 41 %, respectively. Patients who were shock or respiratory failure at the time of ICU admis- admitted due to acute renal failure, cardiac arrest, sep- sion had worse outcomes than patients who lacked these tic shock, or respiratory failure had poor ICU outcomes findings on admission (Supplemental Digital Content— and median survival durations ranging from 1 to 21 days. Tables 1, 2). The median SOFA scores at the times of admission and The median SOFA scores on admission were 6 (IQR discharge were significantly lower in ICU survivors than 3.5–9) in ICU survivors and 10 (IQR 6.5–14) in non-sur- in non-survivors. An increase in SOFA score during vivors. In addition, the median maximum SOFA scores the ICU stay is an important predictor of poor survival were 7 (IQR 4–9.5) in survivors and 14 (IQR 10–17) in outcomes. We also determined that a higher number of non-survivors, and the median discharge SOFA scores organ failures was associated with an increased risk of were 4 (IQR 2–6) in survivors and 10 (IQR, 7.5–13.5) in ICU mortality (Supplemental Digital Content—Table 1). Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 4 of 12

Table 1 Patient and clinical characteristics by Alive or Death at the Intensive Care Unit (ICU) All patientsa Alive at ICU dischargeb Death at ICUb p* (N 172) (N 132) (N 40) = = = Age (y) Median (IQR) 52 (38–62) 52 (38–62) 53.5(36–66) 0.65 Gender Female 93 (54.1 %) 71 (76.3 %) 22 (23.7 %) Male 79 (45.9 %) 61 (77.2 %) 18 (22.8 %) 0.89† Histological diagnosis Unclassified high-grade sarcoma 43 (25.0 %) 36 (83.7 %) 7 (16.3 %) Bone sarcomac 30 (17.4 %) 21 (70 %) 9 (30 %) Vasculard 17 (9.9 %) 16 (94.1 %) 1 (5.9 %) GISTe 11 (6.4 %) 10 (90.9 %) 1 (9.1 %) MFHe 11 (6.4 %) 6 (54.5 %) 5 (45.5 %) Muscle 10 (5.8 %) 7 (70 %) 3 (30 %) Leiomyosarcoma 13 (7.6 %) 12 (92.3 %) 1 (7.7 %) Liposarcoma 7 (4.1 %) 4 (57.1 %) 3 (42.9 %) Synovial sarcoma 9 (5.2 %) 5 (55.6 %) 4 (44.4 %) Others 21 (12.2 %) 15 (71.4 %) 6 (28.6 %) 0.09 Status of malignancy First course of chemotherapy 29 (16.9 %) 20 (69 %) 9 (31 %) Progression 59 (34.3 %) 45 (76.3 %) 14 (23.7 %) Stable disease or partial remission 61 (35.5 %) 50 (82 %) 11 (18 %) Complete remission 4 (2.3 %) 4 (100 %) 0 (0 %) Mixed response 6 (3.5 %) 4 (66.7 %) 2 (33.3 %) Unknown 13 (7.6 %) 9 (69.2 %) 4 (30.8 %) 0.57 Site of malignancy Head and neck 15 (8.7 %) 11 (73.3 %) 4 (26.7 %) Thoracic 43 (25 %) 34 (79.1 %) 9 (20.9 %) Abdomen 71 (41.3 %) 55 (77.5 %) 16 (22.5 %) Extremities 43 (25 %) 32 (74.4 %) 11 (25.6 %) 0.94 Organ metastasis Lung No 86 (50 %) 69 (80.2 %) 17 (19.8 %) Yes 86 (50 %) 63 (73.3 %) 23 (26.7 %) 0.28 Liver No 142 (82.6 %) 108 (76.1 %) 34 (23.9 %) Yes 30 (17.4 %) 24 (80 %) 6 (20 %) 0.64 Other No 79 (45.9 %) 57 (72.2 %) 22 (27.8 %) Yes 93 (54.1 %) 75 (80.6 %) 18 (19.4 %) 0.19 Number of organ metastasis 0 42 (24.4 %) 32 (76.2 %) 10 (23.8 %) 1 54 (31.4 %) 39 (72.2 %) 15 (27.8 %) 2 76 (44.2 %) 61 (80.3 %) 15 (19.7 %) 0.56 ≥ Treatment Current chemotherapy regimen None 43 (25 %) 34 (79.1 %) 9 (20.9 %) Adriamycin-based 68 (39.5 %) 50 (73.5 %) 18 (26.5 %) Gemcitabine-based 20 (11.6 %) 13 (65 %) 7 (35 %) Targeted therapy 41 (23.8 %) 35 (85.4 %) 6 (14.6 %) 0.29 Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 5 of 12

Table 1 continued All patientsa Alive at ICU dischargeb Death at ICUb p* (N 172) (N 132) (N 40) = = = No. cycles of current chemotherapy, median 1 (0–3) 1 (0–3) 1 (1–2) 0.80† (IQRe) No. cycles of prior chemotherapies, median (IQR) 1 (0–3) 1 (0–3) 0 (0–2) 0.037† No. cycles of prior chemotherapies 0–1 114 (66.3 %) 88 (77.2 %) 26 (22.8 %) 2 58 (33.7 %) 44 (75.9 %) 14 (24.1 %) 0.85 ≥ Radiation No 109 (63.7 %) 80 (73.4 %) 29 (26.6 %) Yes 62 (36.3 %) 51 (82.3 %) 11 (17.7 %) 0.19 Clinical conditions present at ICU admission Anemia 144 (83.7 %) 110 (76.4 %) 34 (23.6 %) 0.80 Hypotension 92 (53.5 %) 63 (68.5 %) 29 (31.5 %) 0.006 Septic shock 53 (30.8 %) 33 (62.3 %) 20 (37.7 %) 0.003 Bacteremia 21 (12.2 %) 14 (66.7 %) 7 (33.3 %) 0.24 Thrombocytopenia 86 (5 %) 68 (79.1 %) 18 (20.9 %) 0.47 Respiratory failure 74 (43 %) 39 (52.7 %) 35 (47.3 %) <0.0001 Acidosis 70 (40.7 %) 42 (60 %) 28 (40 %) <0.0001 Altered mental status 65 (37 %) 42 (64.6 %) 23 (35.4 %) 0.003 Abnormal glucose 60 (34.9 %) 42 (70 %) 18 (30 %) 0.13 Acute renal failure 58 (33.7 %) 35 (60.3 %) 23 (39.7 %) 0.0003 Pancytopenia 58 (33.7 %) 47 (81 %) 11 (19 %) 0.34 Pneumonia 51 (29.7 %) 32 (62.7 %) 19 (37.3 %) 0.005 Neutropenia (ANCe <1500/mm3) 51 (29.7 %) 40 (78.4 %) 11 (21.6 %) 0.73 Cardiac dysrhythmia 35 (20.3 %) 23 (65.7 %) 12 (34.3 %) 0.08 Heart failure 32 (18.6 %) 21 (65.6 %) 11 (34.4 %) 0.10 Hypertension 25 (14.5 %) 22 (88 %) 3 (12 %) 0.20* Malnutrition (protein/calorie) NOS 23 (13.4 %) 15 (65.2 %) 8 (34.8 %) 0.16 Hemorrhage 15 (8.7 %) 12 (80 %) 3 (20 %) 1.0* Gastrointestinal hemorrhage 13 (7.6 %) 12 (92.3 %) 1 (7.7 %) 0.30* Cardiac arrest 11 (6.4 %) 3 (27.3 %) 8 (72.7 %) 0.0004* Pulmonary embolism 11 (6.4 %) 8 (72.7 %) 3 (27.3 %) 0.72* Atrial fibrillation 10 (5.8 %) 7 (70 %) 3 (30 %) 0.70* Seizures/convulsions 9 (5.2 %) 8 (88.9 %) 1 (11.1 %) 0.69* ICU admission data Mechanical v entilator 75 (43 6 %) 40 (30.3 %) 35 (87.5 %) <0.0001 · Charlson comorbidity index, median (IQR) 6 (6–7) 6 (6–7) 6 (4.5–7) 0.70† 2 26 (15.1 %) 19 (73.1 %) 7 (26.9 %) ≤ >2 146 (84.9 %) 113 (77.4 %) 33 (22.6 %) 0.63 SOFA admission score, median (IQR) 7 (4–10) 6 (3.5–9) 10 (6.5–14) <0.0001† Max SOFA admission score, median (IQR) 8 (5–12) 7 (4–9.5) 14 (10–17) <0.0001† SOFA discharge score, median (IQR) 5 (3–8) 4 (2–6) 10 (7.5–13.5) <0.0001† No. organ failures, median (IQR) 1 (0–2) 1 (0–2) 2 (1–3) <0.0001

* Fisher exact p value † Wilcoxon rank-sum test a Count (column %—percent of admissions with that variable) are presented unless specified b Count (row %—percent of N in column with All patients) are presented unless specified c Ewing sarcoma, osteosarcoma, chondrosarcoma d Angiosarcoma, epithelioid hemangioendothelioma e Gastrointestinal stromal tumor, Malignant fibrous histiocytoma, Interquartile range, Sequential Organ Failure Assessment, Absolute Neutrophil Count Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 6 of 12

Table 2 Univariate logistic regression model for death at ICU Odds Ratio 95 % CI p

Reason for ICU admission acidosis: yes v. no 5.000 2.317 10.788 <0.0001 Reason for ICU admission renal failure: yes v. no 3.750 1.795 7.831 0.0004 Reason for ICU admission cardiac arrest: yes v. no 10.750 2.699 42.824 0.0008 Reason for ICU admission cardiac dysrhythmia: yes v. no 2.031 0.902 4.575 0.0873 Reason for ICU admission hypotension: yes v. no 2.887 1.332 6.258 0.0072 Reason for ICU admission pneumonia: yes v. no 2.827 1.353 5.910 0.0057 Reason for ICU admission respiratory failure: yes v. no 16.691 6.086 45.775 <0.0001 Reason for ICU admission septic shock: yes v. no 3.000 1.439 6.253 0.0034 Number of organ failures 1.933 1.417 2.637 <0.0001 Number of mets 0.997 0.745 1.334 0.9835 Histology diagnosis Unclassified high grade sarcoma 1.000 Bone (Ewing, Osteo, Chondrosarcoma) 2.204 0.716 6.788 0.1685 Vascular 0.321 0.036 2.833 0.3068 GIST 0.514 0.056 4.685 0.5552 MFH 4.286 1.019 18.029 0.0471 Muscle 2.204 0.456 10.661 0.3258 Leiomyosarcoma 0.429 0.048 3.848 0.4493 Liposarcoma 3.857 0.703 21.153 0.1200 Synovial sarcoma 4.114 0.878 19.270 0.0726 Others 2.057 0.592 7.149 0.2564

Table 3 Trends in Sequential Organ Failure Assessment (SOFA) scores and their impact on ICU and overall survival Increase in SOFA score during ICU No. patients in Patients surviving ICU Patients surviving till end Median survival, stay (Maximum SOFA- Admission subgroup stay, % of study, % d (range) SOFA)

SOFA scores during ICU stay 0 or less 103 86.4 22.3 187 (108–347) 1–3 43 81.4 27.9 96 (21–193) 4–5 10 60.0 10.0 12 (2–85) 6 15 13.3 6.7 17 (6–22) ≥ Admission SOFA scores 0–4 50 88.0 26.0 159 (49–289) 5–7 46 84.8 21.7 182 (38–347) 8–10 41 80.5 24.4 185 (58–494) 11 35 45.7 11.4 6 (2–19) ≥ Change in SOFA scores of 35 patients with admission SOFA scores of 11 or more 0 18 44.4 5.6 4.5 (0.5–9) ≤ 1 6 83.3 33.3 56 (4–NE) 2 3 66.7 33.3 13 (2–NE) 3 8 12.5 0.0 6 (1–46) ≥ ADM SOFA SOFA score at the time of ICU admission, MAXOFA maximum SOFA score, NE not estimated

The ICU mortality rates observed in the study popula- poor outcomes. Our data have shown that adequate car- tion were consistent with those in the existing medical lit- diac, renal, and respiratory functions play a key role in erature and promote a case for a higher level of care with acute survival. Serial SOFA scores may serve as an objec- aggressive monitoring in patients who lack predictors of tive measure for short-term and long-term prognosis for Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 7 of 12

Table 4 Overall survival estimates by patients and clinical characteristics in percentages No of patients No of deaths 1-month overall 2-month overall 6-month overall p survival estimate survival estimate survival estimate (95 % CI) (95 % CI) (95 % CI)

All patients 172 135 64 (57–71) 57 (49–64) 41 (33–48) Age (years) <65 139 109 64 (56–72) 56 (47–64) 41 (33–50) 65 33 26 64 (45–77) 61 (42–75) 37 (21–54) 0.91 ≥ Gender Female 93 72 66 (56–75) 60 (49–69) 45 (34–55) Male 79 63 62 (50–71) 48 (37–59) 36 (25–47) 0.33 Histological diagnosis Unclassified high-grade 43 34 67 (51–79) 58 (42–71) 46 (30–60) sarcoma Bonesa 30 25 70 (50–83) 60 (40–75) 38 (21–55) Vascularb 17 16 64 (36–82) 51 (25–72) 32 (12–54) GIST 11 7 73 (37–90) 64 (30–85) 45 (17–71) MFH 11 8 55 (23–78) 55 (23–78) 36 (11–63) Muscle 10 9 60 (25–83) 60 (25–83) 20 (3–47) Leiomyosarcoma 13 7 84 (50–96) 84 (50–96) 75 (40–91) Liposarcoma 7 7 29 (4–61) 29 (4–61) 29 (4–61) Synovial sarcoma 9 6 56 (20–80) 56 (20–80) 56 (20–80) Others 21 16 56 (33–74) 45 (23–65) 28 (11–49) 0.07 Status of malignancy First course of chemo- 29 22 66 (45–8) 55 (35–71) 41 (22–59) therapy Progression 59 56 49 (36–61) 37 (25–49) 17 (9–27) Stable disease or partial 61 41 75 (62–84) 72 (58–81) 61 (47–72) remission Complete remission 4 2 100 100 75 (13–96) Mixed response 6 5 67 (19–90) 67 (19–90) 67 (19–90) Unknown 13 9 68 (36–87) 68 (36–87) 29 (7–56) <0.0001 Site of malignancy Head and neck 15 11 52 (25–73) 52 (25–73) 29 (9–53) Thoracic 43 33 67 (51–79) 60 (43–73) 46 (30–60) Abdomen 71 57 62 (49–72) 53 (41–64) 37 (26–48) Extremities 43 34 7 (54–81) 63 (47–75) 46 (31–60) 0.86 Organ metastasis Lung No 86 62 66 (55–75) 59 (48–69) 46 (35–56) Yes 86 73 62 (51–72) 55 (43–65) 36 (25–46) 0.02 Liver No 142 109 66 (58–73) 59 (51–67) 46 (37–54) Yes 30 26 56 (36–72) 45 (27–62) 17 (6–33) 0.0093 Other No 79 55 66 (54–75) 63 (51–73) 49 (38–60) Yes 93 80 63 (52–72) 52 (41–62) 33 (24–43) 0.0091 Number of organ metastases 0 42 25 71 (55–83) 69 (53–81) 59 (43–72) 1 54 43 59 (45–71) 53 (39–65) 40 (26–53) 2 76 67 64 (52–74) 53 (41–64) 31 (21–42) 0.001 + Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 8 of 12

Table 4 continued No of patients No of deaths 1-month overall 2-month overall 6-month overall p survival estimate survival estimate survival estimate (95 % CI) (95 % CI) (95 % CI)

Localized disease No 130 110 62 (53–7) 53 (44–61) 34 (26–43) Yes 42 25 71 (55–83) 69 (53–81) 59 (43–72) <0.0001 Treatment Current chemotherapy regimen None 43 31 57 (41–7) 52 (36–66) 39 (24–53) Adriamycin-based 68 51 69 (56–78) 66 (53–76) 51 (38–63) chemotherapy Gemcitabine-based 20 18 5 (27–69) 45 (23–65) 25 (9–45) therapy Targeted therapy 41 35 71 (54–82) 54 (37–67) 34 (20–49) 0.31 Radiation No 109 85 63 (54–72) 59 (49–67) 39 (29–48) Yes 62 50 65 (52–76) 53 (40–65) 43 (30–55) 0.86 Clinical conditions present at ICU admission Anemia 144 113 66 (57–73) 59 (50–66) 43 (35–51) 0.60 Hypotension 92 73 57 (46–66) 51 (40–61) 42 (32–52) 0.96 Septic shock 53 44 49 (35–62) 45 (32–58) 33 (21–46) 0.10 Bacteremia 21 17 57 (34–75) 57 (34–75) 43 (22–62) 0.58 Thrombocytopenia 86 63 71 (60–79) 64 (52–73) 49 (37–59) 0.037 Respiratory failure 74 62 46 (34–56) 41 (30–52) 26 (16–37) <0.0001 Acidosis 70 58 47 (35–58) 41 (30–53) 35 (24–46) 0.011 Altered mental status 65 57 52 (39–63) 42 (30–54) 25 (15–36) 0.001 Abnormal glucose 60 50 60 (46–71) 49 (36–61) 25 (14–37) 0.017 Acute renal failure 58 49 48 (35–60) 40 (27–52) 29 (18–41) 0.016 Pancytopenia 58 41 79 (66–88) 74 (61–83) 57 (43–69) 0.006 Respiratory abnormality 53 38 73 (59–83) 64 (49–75) 49 (35–62) 0.036 Pneumonia 51 42 53 (38–65) 49 (35–62) 35 (22–48) 0.41 Neutropenia 51 36 78 (64–87) 70 (56–81) 58 (43–70) 0.013 (ANC <1500/mm3) Cardiac dysrhythmia 35 26 57 (39–72) 51 (34–66) 46 (29–61) 0.96 Heart failure 32 25 63 (44–77) 56 (37–71) 53 (34–68) 0.80 Hypertension 25 15 80 (58–91) 71 (48–85) 61 (39–78) 0.049 Malnutrition (protein/ 23 21 60 (37–77) 60 (37–77) 28 (11–47) 0.047 calorie) NOS Hemorrhage (non–gas- 15 9 64 (34–83) 64 (34–83) 47 (19–71) 0.44 trointestinal) Other pulmonary insuf- 14 11 71 (41–88) 64 (34–83) 64 (34–83) 0.71 ficiency Gastrointestinal hemor- 13 9 62 (31–82) 54 (25–76) 38 (14–63) 0.51 rhage Cardiac arrest 11 8 27 (7–54) 27 (7–54) 27 (7–54) 0.17 Pulmonary embolism 11 10 73 (37–90) 63 (29–84) 32 (8–59) 0.21 Atrial fibrillation 10 7 70 (33–89) 70 (33–89) 6 (25–83) 0.40 Seizures 9 8 67 (28–88) 44 (14–72) 30 (5–61) 0.54 ICU admission data Charlson comorbidity index 2 26 15 69 (48–83) 69 (48–83) 54 (33–71) ≤ >2 146 120 63 (55–71) 55 (46–62) 38 (30–46) 0.0031 Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 9 of 12

Table 4 continued No of patients No of deaths 1-month overall 2-month overall 6-month overall p survival estimate survival estimate survival estimate (95 % CI) (95 % CI) (95 % CI)

Charlson comorbidity index <6 42 25 71 (55–83) 69 (53–81) 59 (43–72) 6 130 110 62 (53–70) 53 (44–61) 34 (26–43) <0.0001 ≥ SOFA maximum score <8 79 60 82 (72–89) 71 (60–80) 49 (38–60) 8 93 75 49 (39–59) 45 (34–55) 33 (24–43) 0.005 ≥ SOFA discharge score <5 77 59 80 (69–88) 71 (59–80) 47 (35–58) 5 95 76 51 (41–61) 46 (36–56) 36 (26–45) 0.12 ≥ No. organ failures 1 59 47 76 (63–85) 69 (56–79) 53 (40–65) 2 113 88 58 (48–66) 50 (41–59) 34 (25–43) 0.12 ≥ ADM SOFA SOFA score at the time of ICU admission, MAXOFA maximum SOFA score, NE not estimated a Ewing sarcoma, osteosarcoma, chondrosarcoma b Angiosarcoma, epithelioid hemangioendothelioma

Fig. 1 Kaplan-Meier estimates of overall survival Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 10 of 12

Table 5 Multivariable cox proportional hazards model ICU mortality. Among mortality statistics reported for for overall survival specific tumors, patients with head and neck cancer, lung Variable Hazard 95 % CI p cancer, and gynecological malignancies have been shown ratio P to have ICU mortality rates of 39, 36, and 17.3 %, respec- Histological diagnosis 0.006 tively [11–13]. In contrast, patients with hematological Unclassified high-grade 1.000 cancers have been reported to have higher ICU mortal- sarcoma ity rates: 48.3 and 56 % in two separate studies [14, 15]. Bonesa 0.881 0.516 1.506 0.64 These higher rates may be attributed to increased sever- Vascularb 1.657 0.862 3.185 0.13 ity of illness and higher incidence of sepsis due to associ- GIST 0.281 0.119 0.662 0.004 ated leukopenia in these patients [10]. MFH 0.674 0.236 1.929 0.46 Several ICU scores; including MODS, APACHE II, Muscle 2.048 0.917 4.573 0.08 SAPS II, and SOFA, have been used as objective ways to Leiomyosarcoma 0.375 0.160 0.880 0.02 describe mild to severe organ dysfunction [16–19]. In our Liposarcoma 1.412 0.575 3.468 0.45 study, we used SOFA scores, which have been calibrated Synovial sarcoma 0.793 0.293 2.143 0.65 to predict ICU and in-hospital mortality rates in cancer Others 0.746 0.373 1.493 0.41 patients by evaluating the combination of clinical condi- Status of malignancy 0.003 tions that lead to ICU admission [20–23]. Complete remission 1.000 We found out those patients who had high SOFA scores First course of chemotherapy 2.323 0.473 11.425 0.30 on admission did worse than patients with lower SOFA Mixed response 2.720 0.444 16.677 0.28 scores. Another important objective finding was that Progression 5.172 1.075 24.874 0.04 patients whose SOFA scores continued to rise following Stable disease 2.262 0.501 10.218 0.29 admission to the ICU had the worst outcomes, making Reason for ICU admission pancytopenia these patients appropriate candidates for early supportive No and palliative care. As a general guide, increase in scores Yes 0.351 0.226 0.546 <0.0001 of 6 or more since admission led to drastic changes in Reason for ICU admission respiratory failure outcome and ICU survival decreased from 60 to 13.3 % No (Table 3). A patient who is admitted with a SOFA score Yes 1.703 1.125 2.579 01 of 11 or higher and their score increases by 3 or more has CCI (continuous) 1.139 1.023 1.268 02 dismal ICU survival of 12.5 %. Such a patient should be SOFA score at discharge (con- 1.210 1.141 1.283 <0.0001 considered for transition to hospice (Table 3). tinuous) Among individual clinical findings on ICU admissions, Initially included in the model and then reduced by stepwise selection method: patients with cardiac arrest had the worst prognosis, with histology diagnosis, status of malignancy, number of organ metastasis, number an ICU mortality rate of 72 % (8/11) and a median OS of organ failures, and reason for ICU admission: acidosis, acute renal failure, pancytopenia, respiratory failure, and septic shock. There were 13 patients of 24 h. However, cardiopulmonary resuscitation has without known malignancy status who were not included in this analysis been shown to be a non-beneficial intervention in more CI confidence interval, GIST Gastrointestinal stromal tumor, MFH Malignant than 90 % of the patients with cancer [24–26]. In our fibrous histiocytoma, CCI Charlson comorbidities index, SOFA Sequential Organ univariate analysis, patients with respiratory failure had Failure Assessment a Ewing sarcoma, osteosarcoma, chondrosarcoma worse outcomes than patients with good respiratory sta- b Angiosarcoma, epithelioid hemangioendothelioma tus. Mechanical ventilation has been shown to be associated with increased mortality rates (73 %) in cancer patients, especially those with disseminated disease and sarcoma. There appears to be significant heterogeneity in poor performance status at the time of ICU admission ICU mortality among all cancer patients because of a lack [27]. Similar findings have been observed in patients with of data comparing similar histologies. Our patient popu- hematological, lung, and head and neck cancers in the lation had an ICU mortality rate of 23 %, which is com- ICU setting (64–74 %) [20, 28–30]. parable to the 20 % mortality in solid tumors [10]. The Hypotension and septic shock are additional clinical same study reported that ICU and in-hospital mortality findings that correlated with higher ICU mortality in our rates for cancer patients were not significantly different multivariate analysis. Sepsis is one of the leading causes from those of non-cancer patients [10]. However, among of ICU admission in cancer patients. However, patients patients with multiple organ failures; mortality rates were with infections leading to shock and vasopressor use tend higher in cancer patients than in non-cancer patients. to do poorly and should be managed conservatively [10, This finding is consistent with our study, in which the 13, 15, 27]. Patients with acute kidney injury also have number of organ failures was directly proportional to high acute mortality. This trend is more commonly seen Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 11 of 12

in patients with hematological malignancies [31]. Fur- suggest that patients with advanced malignancy that thermore, renal replacement therapy has been associated are admitted to the ICU for respiratory failure, cardiac with high ICU mortality [24]. Given their bleak progno- arrest, septic shock, acute renal failure or acidosis and ses, patients with these individual or combinations of have high SOFA score with subsequent worsening in the clinical findings should be triaged early so that inappro- ICU have very poor prognosis. Based on the retrospec- priate use of aggressive therapy can be avoided. tive data which needs further validation we can recom- There are limited data evaluating long-term survival mend that judicious approach should be taken in patients rates of critically ill cancer patients after their discharge with predictors of poor survival before subjecting them from the ICU. Six-month OS in sarcoma patients was to aggressive treatment. 41 % in our study, which is comparable to the 6-month Authors’ contributions mortality rates of 59.3 and 66 % reported in critically ill RG, DA, RB, SP, JL, and VR were responsible for the study concept. All authors patients with hematological malignancies [28, 30]. How- contributed to the study design. RG, NR, and EJR collected and assembled these data. NR, JS, and XL analyzed and interpreted these data. RG and NH ever, long-term mortality rates of 63 to 98 % have been wrote the initial draft, and all the authors approved the final draft of the report. reported in patients with lung cancer. The major pre- All authors read and approved the final manuscript. dictors of long-term prognosis in these patients were Author details dependence on mechanical ventilation during the ICU 1 The University of Texas at Houston Internal Medicine Residency Program, stay and progression of cancer after discharge from the Houston, TX, USA. 2 The University of Texas Graduate School of Biomedical ICU [11]. In our study, we found that the status of malig- Sciences at Houston, Houston, TX, USA. 3 Division of Quantitative Sciences, The University of Texas MD Anderson Cancer Center, Houston, TX, USA. 4 Depart- nancy, metastatic disease, Charlson comorbidity index ment of Critical Care, The University of Texas MD Anderson Cancer Center, and SOFA scores at discharge, and the presence of res- Houston, TX, USA. 5 Department of Sarcoma Medical Oncology, The University piratory failure or cytopenia at the time of ICU admis- of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd # 450, Houston, TX 77030, USA. sion were significant predictors of survival outcomes in both univariate and multivariate analyses (Table 4). Acknowledgements Therefore, when looking at patterns of long-term sur- We did not have any financial support. We thank critical care team for providing the initial data and SOFA scores. We thank statisticians Naveen Ramesh, vival after ICU stay, the above-mentioned clinical prog- Xiudong Lei and Juhee Song. The statistical analysis work was supported in nostic factors should be considered early in the clinical part by the Cancer Center Support Grant (NCI Grant P30 CA016672). We thank course. scientific editors, Bryan Tutt and Tamara Locke, in Department of Scientific Publications at MD Anderson for manuscript editing. There were a number of limitations to this study. First, Poster presented at 2014 American Society of Clinical Oncology National this is a single-institution retrospective study covering a Meeting, Chicago, IL. limited number of years. We are a tertiary care referral Competing interests center for sarcomas, so the results of our study may not The authors declare that they have no competing interests. be applicable to smaller hospitals or low-volume centers. Each hospital has its own ICU admission and discharge Received: 5 March 2016 Accepted: 17 June 2016 policies which may bias the results of a single center study. Second, our study did not have a control group to compare the outcomes of critically ill sarcoma patients who were not admitted to the ICU and were managed References conservatively. Even though it is a large collection of a 1. Fletcher CDM, Bridge JA, Hogendoorn P, Mertens F. World Health Organi- zation classification of tumors of soft tissue and bone. 4th ed. Lyon: IARC rare tumor our study outcomes were limited to only 40 Press; 2013. events in the ICU. Due to a small number of events and 2. Mackall CL, Meltzer PS, Helman LJ. Focus on sarcomas. Cancer Cell. multiple risk factors, it is difficult to make a generalized 2002;2:175–8. 3. Chassin MR. Costs and outcomes of medical intensive care. Med Care. presumption of which risk factor individually impacted 1982;20:165–79. the short term survival. In addition, quality-of-life meas- 4. Turnbull A, Goldiner P, Silverman D, et al. The role of an intensive care unit ures and performance status scores could not be reliably in a cancer center. An analysis of 1035 critically ill patients treated for life- threatening complications editorial. Cancer. 1976;37:82–4. collected in the retrospective setting. 5. Shapira DV, Studnicki J, Bradham DD. Intensive care, survival and expense In conclusion, our study is the largest study to date of of treating critically ill cancer patients. JAMA. 1993;269:783–6. OS in sarcoma patients admitted to the ICU. In our study 6. Azoulay E, Moreau D, Alberti C, et al. Predictors of short-term mortality in critically ill patients with solid malignancies. Intensive Care Med. span of 7 years, we have shown that OS in sarcoma is 2000;26:1817–23. comparable with that of critically ill patients with other 7. Lecuyer L, Chevret S, Thiery G, et al. The ICU trial: a new admission policy solid tumors. The admission SOFA scores and change for cancer patients requiring mechanical ventilation. Critical Care Med. 2007;35:808–14. in the SOFA scores during ICU stay are highly reliable 8. Azoulay E, Soares M, Darmon M, et al. Intensive care of the cancer patient: indicators of probability of survival and should be used recent achievements and remaining challenges. Ann Intensive Care. in decision making in critically ill patients. Our results 2011;1:5. Gupta et al. Clin Sarcoma Res (2016) 6:12 Page 12 of 12

9. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prog- 22. Owczuk R, Wujtewicz MA, Sawicka W, et al. Patients with haematologi- nostic comorbidity in longitudinal studies: development and validation. J cal malignancies requiring invasive mechanical ventilation differences Chronic Dis. 1987;40:373–83. between survivors and non-survivors in intensive care unit. Support Care 10. Taccone FS, Artigas AA, Sprung CL, et al. Caracteristics and outcomes of Cancer. 2005;13:332–8. cancer patients in European ICUs. Crit Care. 2009;13:R15. 23. Cárdenas-Turanzas M, Ensor J, Wakefield C, et al. Cross-validation of a 11. Soubani AO, Ruckdeschel JC. The outcome of medical intensive sequential organ failure assessment score-based model to predict mor- care for lung cancer patients: the case for optimism. J Thorac Oncol. tality in patients with cancer admitted to the intensive care unit. J Crit 2011;6:633–8. Care. 2012;27:673–80. 12. Namendys-Silva SA, González-Herrera MO, Texcocano-Becerra J, et al. 24. Heo SJ, Kim G, Lee CK, et al. Prediction of short- and long-term survival Outcomes of critically ill gynecological cancer patients admitted to for advanced cancer patients after ICU admission. Support Care Cancer. intensive care unit. Am J Hosp Palliat Care. 2013;30:7–11. 2014;1:1–9. 13. Soares M, Salluh JI, Ferreira CG, et al. Prognosis of lung cancer patients 25. Wallace S, Ewer MS, Price KJ, et al. Outcome and cost implications of with life-threatening complications. Chest. 2007;131:840–6. cardiopulmonary resuscitation in the medical intensive care unit of a 14. Pène F, Azoulay E, Blot F, et al. Outcome of critically ill allogeneic hemat- comprehensive cancer center. Support Care Cancer. 2002;10:425–9. opoietic stem-cell transplantation recipients: a reappraisal of indications 26. Reisfield GM, Wallace SK, Munsell MF, et al. Survival in cancer patients for organ failure supports. J Clin Oncol. 2006;24:643–9. undergoing in-hospital cardiopulmonary resuscitation: a meta-analysis. 15. Geerse DA, Span LF, Pinto-Sietsma SJ, et al. Prognosis of patients with Resuscitation. 2006;71:152–60. haematological malignancies admitted to the intensive care unit: 27. Azevedo LC, Caruso P, Silva UV, et al. Outcomes for patients with cancer sequential Organ Failure Assessment (SOFA) trend is a powerful predictor admitted to the ICU requiring ventilatory support: results from a prospec- of mortality. Eur J Intern Med. 2011;22:57–61. tive multicenter study. Chest. 2014;146:257–66. 16. Marshall JC, Cook DJ, Christou NV, et al. Multiple organ dysfunction 28. Bird GT, Farquhar-Smith P, Wigmore T, et al. Outcomes and prognostic fac- score: a reliable descriptor of a complex clinical outcome. Crit Care Med. tors in patients with hematological malignancy admitted to a specialist 1995;23:1638–52. cancer intensive care unit: a 5 yr study. Br J Anesth. 2012;108:452–9. 17. Vincent JL, Moreno R, Takala J, et al. The SOFA (Sepsis-related Organ Fail- 29. Soares M, Salluh JI, Toscano L, et al. Outcomes and prognostic factors in ure Assessment) score to describe organ dysfunction/failure. On behalf of patients with head and neck cancer and severe acute illnesses. Intensive the Working Group on Sepsis-Related Problems of the European Society Care Med. 2007;33:2009–13. of Intensive Care Medicine. Intensive Care Med. 1996;22:707–10. 30. Benoit DD, Vandewoude KH, Decruyenaere JM, et al. Outcome and early 18. Knaus WA, Draper EA, Wagner DP, et al. APACHE II: a severity of disease prognostic indicators in patients with a hematologic malignancy admit- classification system. Crit Care Med. 1985;13:818–29. ted to the intensive care unit for a life-threatening complication. Crit Care 19. Le Gall JR, Lemeshow S, Saulnier F. A new simplified acute physiology Med. 2003;31:104–12. score (SAPS II) based on a European/North American multicenter study. 31. Hampshire PA, Welch CA, McCrossan LA, et al. Admission factors associ- JAMA. 1993;270:2957–63. ated with hospital mortality in patients with haematological malignancy 20. Reichner CA, Thompson JA, O’Brien S, et al. Outcome and code admitted to UK adult, general critical care units: a secondary analysis of status of lung cancer patients admitted to the medical ICU. Chest. the ICNARC Case Mix Programme Database. Crit Care. 2009;13:1–17. 2006;130:719–23. 21. Soares M, Caruso P, Silva E, et al. Characteristics and outcomes of patients with cancer requiring admission to intensive care units: a prospective multicenter study. Crit Care Med. 2010;38:9–15.

Submit your next manuscript to BioMed Central and we will help you at every step: • We accept pre-submission inquiries • Our selector tool helps you to ﬁnd the most relevant journal • We provide round the clock customer support • Convenient online submission • Thorough peer review • Inclusion in PubMed and all major indexing services • Maximum visibility for your research

Submit your manuscript at www.biomedcentral.com/submit